Overview
Watch the tutorial
Integrations connect Leena AI to the external systems your organization uses — HRIS, ITSM, identity providers, collaboration tools, and more. Once connected, these integrations become available in Workflows Studio and Skill Studio, enabling automated workflows and AI Colleagues to interact with your enterprise systems.
All integrations are set up through the Unified Dashboard. No custom development is required.
Note: Knowledge connectors (used to sync content into Leena AI's knowledge base) are managed separately under Knowledge Management settings. This page covers integrations used in Workflows Studio and Skill Studio.
How Integrations Are Used
Once an external system is connected, it can be used in two areas of the platform:
Workflows Studio — Connected systems expose actions that can be used as steps in automated workflows. This allows you to build processes that span multiple systems in a single workflow.
Skill Studio — Integrations surface as MCP (Model Context Protocol) tools that can be assigned to AI Colleague tools. This enables AI Colleagues to retrieve data from connected systems or take actions during employee conversations.
How you enable and configure an integration within Workflows Studio or Skill Studio depends on the specific connector and use case. Refer to the respective documentation for each.
Browsing Integrations
The integration catalog is available at Admin Console → Integrations. It displays the list of available integrations, each showing its name, description, and current connection status.
You can browse the catalog or search by name. The list is updated as new integrations are added to the platform.
Connecting an Integration
To set up an integration, select it from the catalog and follow its setup flow. Each integration has its own authentication requirements and configuration steps — these vary by connector.
Once a connection is established, you can view and manage it from the integration's connection details page.
For setup instructions specific to each connector, see the dedicated guide linked from the integration's detail page.
Managing Connections
Each connected integration has a connection details page where you can view the connection status and manage its configuration. If a connection needs attention — for example, due to a credential issue — you can update it from this page.
Connections & authentication concepts
Every connector runs on top of a connection. The sections below explain the concepts that all connectors share — what a connection is, how authentication works, whose identity the Agent acts under, how access stays valid over time, and how the platform keeps repeated actions safe. Understanding them once lets you set up and govern any integration with confidence, instead of relearning the basics for each one.
Connections
A connection is the authenticated bridge between your Leena AI Agents and a third-party application. It holds everything the platform needs to talk to that system on your behalf: where to send requests, and the credentials that authorize them. Actions, tool discovery, and background data lookups (such as populating dropdowns) all run in the context of a connection.
A connection keeps two kinds of information separate. Connection settings — endpoints, identifiers, and other non-secret configuration — are stored as plain configuration the platform can read and display. Credentials — secrets such as keys, passwords, and tokens — are stored encrypted and are decrypted only at the moment a request is actually made. This separation lets administrators view and manage a connection without ever exposing its secrets.
Each connection also has an enabled state that acts as a master switch. Disabling a connection immediately removes its connector from action execution, tool discovery, and background lookups — without deleting any configuration. This lets you suspend an integration and re-enable it later without rebuilding it.
Authentication methods
When you set up a connection, the connector determines how it proves its identity to the third-party system. Leena AI supports a range of authentication methods so each connector can use whatever its provider expects.
| Method | How it authenticates | Typically used for |
|---|---|---|
| API key | A static key issued by the provider is sent with each request | Simple server-to-server access |
| Basic authentication | A username and password are encoded and sent on each request | Systems that expose basic-auth endpoints |
| Bearer token | A static token is sent in the request's authorization header | Providers that issue long-lived tokens |
| OAuth 2.0 | The platform exchanges credentials for an access token and refreshes it automatically | Modern apps that support delegated, token-based access |
| Service-account OAuth | A trusted system identity (for example, certificate- or assertion-based) obtains tokens without an interactive sign-in | Backend integrations that act as the organization |
| Certificate | A client certificate and private key establish a mutually authenticated connection | Systems that require certificate-based trust |
| No authentication | The connection stores only endpoint configuration and sends no credentials | Open endpoints, or systems where access is handled elsewhere |
Whichever method a connector uses, the same security principle applies: credentials are stored encrypted and decrypted only when a request is made, and they are kept separate from non-secret connection settings.
Admin OAuth vs User OAuth
Authentication also answers an important governance question: when an Agent acts on a connected system, whose identity and permissions does it use? Leena AI supports two models.
Admin OAuth (shared connection). An administrator configures one connection that every Agent and workflow shares. All actions run under that single, centrally governed identity, regardless of which employee triggered them. This is the right fit for system-to-system automation — creating tickets, reading records, posting to a shared channel — where the action should always run with the same access.
User OAuth (per-user authentication). Some scenarios require the Agent to act as the individual employee rather than a shared account — for example, creating a file in the requester's own drive, or taking an action that should respect that person's own permissions. Here, each user authorizes the connection with their own account the first time a workflow needs it. The platform prompts them to sign in, then resumes the workflow automatically. After that first authorization they are not asked again unless their access expires or is revoked. For the complete user-authorization flow, see End-user OAuth on connections.
Whether a connector uses Admin OAuth or requires User OAuth is configured on the connector itself in Integrations, not in the workflow builder. A few connectors — such as Pipedream and custom MCP connectors — let you choose between Admin OAuth (shared) and User OAuth (per-user) to match your use case.
| Admin OAuth | User OAuth | |
|---|---|---|
| Who configures it | An administrator, once | Each employee, on first use |
| Identity actions run under | A single shared service identity | The individual employee's account |
| Permissions applied | The shared connection's access | Each user's own access in the target system |
| User prompted to sign in | No | Yes, the first time a workflow needs it |
| Best suited for | System-to-system automation with consistent, centrally governed access | Actions that must act under each user's own permissions |
OAuth token refresh
OAuth 2.0 connections rely on an access token to authorize each request. These tokens are short-lived by design and expire periodically. Leena AI manages their full lifecycle so connections do not break as tokens expire.
Automatic refresh
Before each action, the platform checks the access token. If it has expired — or is within a short buffer before expiry (five minutes by default) — Leena AI uses the stored refresh token to obtain a new access token, saves it, and continues the request. This happens transparently, with no interruption and no manual intervention.
Refresh works the same way for both Admin OAuth and User OAuth connections; where access is per-user, each employee's tokens are refreshed independently. For this to work, the provider must issue a refresh token, so Leena AI requests long-lived (offline) access wherever the provider supports it. Some providers issue non-expiring tokens, which the platform recognizes and never needs to refresh.
Non-refreshable credentials and expiry reminders
API keys, static bearer tokens, and certificates cannot refresh themselves. For these, Leena AI helps you stay ahead of expiry with credential expiry reminders — configurable email notifications (for example, 30, 60, or 90 days, or a custom window before expiry) sent to chosen recipients. This prompts your team to rotate credentials before a connection stops working, so integrations do not fail silently.
Idempotency
When Agents act on external systems, an action call occasionally has to be retried — because of a network blip, a slow response, or a transient gateway error. Without safeguards, a retry could create the same ticket twice or send a duplicate notification. Leena AI prevents this through idempotency: the guarantee that retrying the same action does not cause it to happen more than once.
How it works
Every action call is tagged with a unique idempotency key. If the platform has to retry that call, it reuses the same key, so both Leena AI and the target system can recognize it as the same request rather than a new one. De-duplication applies for up to 24 hours per key.
Replays and in-progress requests
| Scenario | What happens |
|---|---|
| The same call is retried after it already completed | The original result is returned (replayed); the action is not run again |
| The same call is retried while the first is still in progress | The duplicate is briefly held off and asked to retry, so only one execution proceeds |
| The same key arrives with a different request | It is rejected, preventing one key from being reused for a different operation |
The net effect is that transient failures and automatic retries do not turn into duplicate side effects — no double tickets, duplicate calendar events, or repeated messages — while genuine new requests still go through normally.
Getting Started
- Navigate to Unified Dashboard → Global Settings → Integrations.
- Find the system you want to connect.
- Follow the guided setup for that connector.
- Once connected, refer to the Workflows Studio or Skill Studio documentation to start using the integration.
Updated about 1 month ago
